Social Networks As Reputation Systems For SAFE


#1

I’ve been mulling over how to create reputation systems for SAFE and want to share some initial sketchy ideas. Please let’s not go off topic here (e.g onto whether this is something you want or not, please keep as close to how to build a reputation system as possible, and particularly the merits or problems with the ideas presented here).

Reputation & Relevance

When I talk about reputation, I don’t envisage something about good/bad but as something I want/Don’t want to engage with, and the ability for groups to gravitate around common “interests”, but an emphasis on trust. So I am talking about relevance as well as reputation. I want to still call it a reputation system though, because I don’t want to lose the connection with trust, without which it degenerates into a search engine.

Since I originally started mulling this over I’ve begun to think differently about how such systems could be built.

Originally I was thinking of something using the gamification approach of StackExchange, not necessarily upvotes and downvotes, but in terms of how to get people to engage.

Social Reputation

Lately though, I’ve been thinking that it could be far more seamless, almost invisible. For example, what is twitter, but a very effective reputation system. Not in crude terms like number of followers, but how using it enables us to connect with those people that we are interested in, out of an enormous number of accounts. I rarely search twitter - through using it I gravitate towards people who share stuff that I value, and by following and retweeting them, I enhance their reputation - not by increasing a score, but by increasing their visibility with people already interested in what I share.

If we were to look at the relationships, we would see how this implicit reputation system is encoded by who follows whom, who is retweeted by whom etc., and how we are guided by this invisible reputation network without being conscious of its presence. For example, I follow people very selectively. I see new people who I might follow when those I already follow retweet things they like. Following and retweeting, unfollowing, blocking etc, become the upvotes and downvotes which we don’t perceive directly, but which are effective, and less crude indicators of usefulness, and in some sense, of reputation.

Implementation

I guess for similar social applications these techniques can be used pretty directly. I’m curious though, to see if we can extract and reuse, or perhaps innovate to bring seamless reputation systems to none social apps too. Or should we provide an underlying social platform, that can act as a reputation and relevance system, integrating with apps that can make use of this?


#2

Question: What if you want reputation but aren’t social? What if you just want to put your stuff out there but don’t want to be extremely social about it? Is gaining reputation limited to extraverts?


#3

Thanks for opening up an important line of inquiry @happybeing! I have been pondering distributed trust systems for quite some time (years, actually), and have just recently landed a high-level architecture that I feel has huge potential. This is independent of the MaidSafe context, but I am hopeful that they will play well together.

I’m writing this now for the first time… Consider this a first draft only… Here goes:

Each node in the trust graph sees only those nodes that they explicitly trust (connections), or nodes that their connections trust. This can cascade to an arbitrary number of levels (degree). The amount of trust (strength of the edge connecting the nodes) falls off logarithmically at each degree.

What is this binding force we are calling trust, really, at the implementation level? I suggest it is one thing and one thing only: private sharing of data. Those nodes (people or organizations) that you share information with, in this vision, are your trust network.

This one simple thing is enough. Everything you care about – your own creations, your collaborations, your curated collections of anything, your annotations – these are your works, a kind of digital wealth. When you share these nuggets of wealth privately, you contribute to an ecosystem of value, a distributed omnicentric user-owned walled garden topology.

That is to say, you are creating real value. Your trusted contacts care about what you are sharing. The content is the value. Sharing the content creates edges in your own trust network subgraph. You can map your trust networks and choose which trust-subgraph data to share with which nodes. This trust graph data is another piece of enormously valuable content – notably because it can be stitched together into cascading networks of trust, thereby enriching the trust networks of your connections.

To summarize: the content is the value. The trust network is created implicitly by sharing that value. Which creates more valuable content – an explicit trust graph that can be shared identically to other content.

OK. While it’s great to get that out, even in very rough form, I should say now that I have no idea whether private sharing, upon which all this heavily relies, is even a thing in maidsafe. I’m still absorbing the basics.

In any case, please feel free to take and improve on anything here that you find useful.

Thanks again for asking great questions!


#4

@Blindsite2k:

Question: What if you want reputation but aren’t social? What if you just want to put your stuff out there but don’t want to be extremely social about it? Is gaining reputation limited to extraverts?

I’m not sure this is a problem, at least only a much as it is in real life, in which case you have to rely on trust signals from other sources than your network and associations. But such “badges” of trust (e.g. membership of industry bodies, regulators etc) all carry centralisation risk, which is what we’re trying to tackle here, so I think I see SAFE as enabling even introverts (I am one by the way), to obtain the benefits of social trust networks without having to go outside and talk to people [shiver :-)].

I see the idea I outlined as tying to model the ways humans behave, how they develop that, and how trust affects their behavior, but extending it way way beyond the 170 people we can remember limit that we hit, and lead to the centralization of trust following the agricultural revolution (or whatever it was!) - ref: The Four Pillars Of A Decentralised Society - TEDx talk by Johan Gevers

@Harlan thanks for sharing your idea which you won’t be surprised to hear, sounds very interesting to me. I like that you are keeping it simple, where I like to leap in and complicate things (wanting to include context and relevance, as well as trust!) I would definitely like to see your idea develop, and try to get something working on SAFE, add well as trying to understand mite myself about humans natural approach to trust and how we apply it in different contexts